AMD ROCm Finance documentation#
2025-12-15
1 min read time
Within the ROCm ecosystem, the AMD ROCm Finance toolkit (ROCm-Finance) includes production‑ready support for industry‑leading Gradient Boosting Machine (GBM) libraries for high-performance finance use cases: microsoft/LightGBM and Xtra-Computing/ThunderGBM.
LightGBM |
ThunderGBM |
|
|---|---|---|
Finance use cases |
|
|
Best for |
Large datasets |
GPU-intensive tasks |
Tree growth strategy |
Leaf-wise |
Level-wise (GPU) |
Categorical feature handling |
Manual + binning |
Manual encoding |
Overfitting control |
L1/L2 + sampling |
Regularization |
These libraries are accelerated on AMD Instinct GPUs through optimized ROCm libraries, leveraging the latest AMD GPU architectures. ROCm enablement of these libraries introduces optimized kernels, memory‑management enhancements, and seamless multi‑GPU scaling, delivering performance gains over CPU-only baselines in intensive workloads.
The ROCm-Finance source code is hosted on GitHub at ROCm/ROCm-Finance.
ROCm-Finance documentation is organized into the following categories: